Documentation

Import MRF-CRBLoss as:

import MRF

Settings

Training_parameters	Define the hyperparameters to train the neural network.
Loss	Define the type of loss function used to train the neural network ON A SPECIFIC parameter.
Normalization	Define the way you want to normalize (or not) the input signal.
Optimizer	Select the optimizer used for the training.
NoiseType	Describe the way you want to define the noise.
Initialization	Specify the way to deal with the first linear layer if your network starts with a projection of the signal onto a low dimensional subspace.
Sampling	Define the way you want to sample the parameters.

Model (Architecture)

BaseModel	Each new network should inherit from this class and be saved in the folder ‘models’.
Projection	Class gathering the parameters to characterize the first linear layer which performs the projection of the input signal onto a low dimensional subspace (if the chosen architecture includes a projection).

Data Preprocessing

BaseData_class	Children classes define the way the fingerprints are generated.
Offline.Data_class	Class allowing to deal with the importation of the precomputed fingerprints.

Network training

BaseNetwork	Mother of Network classes that would contains the method for training.
Offline.Network	Class defining the whole neural network for training.

Validation data and Visualization Tools

Offline.Performances

Class designed to handle the computations and the definition of the validation loss and errors.

Data Loading

Offline.loading_data.loading_data

Installation Tutorials